- Top of page
- Supporting Information
Repeatable quantitative methods for identifying and delineating sensitive habitats are an essential prerequisite for the effective management of human activities (Bax & Williams 2001). Knowledge of sensitivity can be used to assess the sustainability of existing impacts, develop spatial management plans and support environmental impact assessment. Despite the need to identify and delineate sensitive habitats, available methods usually rely on expert judgement and/or scoring systems and are neither repeatable nor verifiable (Johnson & Gillingham 2004; Zacharias & Gregr 2005). This may not be an issue when comparing fundamentally different types of habitat, such as coral reefs and mobile sand in the marine environment (Hall-Spencer, Allain & Fossa 2002) or forest and grassland, but it is an issue when dealing with the full continuum of habitat types across subtle environmental gradients and their associated sensitivities to different impacts.
A suitable definition of sensitivity in the present context is that of Zacharias & Gregr (2005): the degree to which features of the environment (in the current context habitats) respond to stresses, where stresses are deviations of environmental conditions beyond the expected range. If sensitivity is to be measured and reported in a comparable way, and as habitat types and sources of stress are numerous in most environments, a widely applicable scale for measuring and reporting sensitivity is needed. One option is to treat sensitivity as the inverse of recovery time following a stress of defined magnitude and duration. This builds on the conceptual approach developed by Bax & Williams (2001), where sensitivity is highest when resistance and resilience are low. Resistance is defined as the initial resistance of a habitat to disturbance, and resilience is defined to be inversely proportional to recovery time.
In this study, we developed a quantitative, repeatable and validated method for estimating the sensitivity of marine habitats to stress. The method can be used to produce maps of habitat sensitivity for the marine environment. Consistent with Zacharias & Gregr (2005), our approach takes account of the effects of the range of natural stresses to which habitats are exposed. The method we present classifies the sensitivity of seabed habitats in relation to bottom-trawling disturbance. Direct changes in the relative abundances of species killed by fishing gears comprise the most widespread direct human impact on marine ecosystems. Disturbance caused by bottom-trawling is among the most widespread human impact on marine ecosystems, with >50% of many shelf seabeds impacted annually (Watling & Norse 1998; Hall 2002).
The method we present could readily be modified to classify sensitivity to other types of physical disturbance (e.g. aggregate dredging and cable laying) for which the response and recovery of biological attributes (e.g. biomass, diversity and production) have been ascertained. In qualitative terms, the sensitivities of seabed habitat to trawling disturbance are well known. Sensitive habitats are typically found in environments where there are low levels of natural disturbance resulting from wave erosion, currents at the seabed and temperature fluctuation (among others). Such habitats are characterized by the occurrence of large and old individuals (biota) that tend to be relatively abundant (Hall 1999). The response of a sensitive habitat to a given frequency and magnitude of disturbance is characterized by larger reductions in biomass, production and species richness of the associated fauna than those seen in a less sensitive habitat (Hiddink et al. 2006b).
Various conceptual models of the relationships between habitat sensitivity to trawling and habitat type have been proposed (Auster 1998; Jennings & Kaiser 1998) but these have limited application in management decision making because they do not provide quantitative predictions and are therefore not suitable for monitoring performance in relation to objectives or assessing the effects of mitigation measures.
With the advent of an ecosystem approach to fisheries (EAF) (Sinclair & Valdimarsson 2003), it is necessary to develop quantitative methods for predicting sensitivity to external forcing. This is because the EAF requires that managers take account of the ecosystem effects of fishing in management plans that are also intended to achieve sustainable exploitation of target species (Kaiser et al. 2002). Knowledge of habitat sensitivity would therefore support assessment and management. For example, methods for assessing and predicting the relative ecological impacts (where impact is defined as the reduction of some ecological function, such as production) of trawling disturbance in habitats with different sensitivities might provide a rational basis for assigning fishing access rights and identifying undesirable interactions between fishing and the environment. Moreover, the permissible spatial distribution and intensity of trawling on different habitats could be predicated on knowledge of sensitivity, and/or charges for access to fishing rights in different areas could be linked to measures of habitat sensitivity. Such processes might support the development of habitat quotas that are set to maintain a target habitat ‘stock’ (Holland & Schnier 2006). Estimates of sensitivity would also allow the production of habitat sensitivity maps. These could contribute to environmental impact assessment, if this process was required for new fisheries, by helping to identify areas where trawling would be expected to have the greatest environmental impact. Knowledge of the distribution of fish populations and the sensitivity of their habitats could also be used to redirect trawling to more resilient areas.
We developed a method for assessing the sensitivity of seabed habitats to bottom-trawling disturbance. The method assumes that sensitivity is related to the recovery time of biomass or production, as predicted using a validated size-based model that takes account of the effects of natural disturbance on habitat characteristics. The method was applied to seabed habitats in the North Sea and used to delineate and map habitat sensitivity at large spatial scales (> 105 km2). Although we used the North Sea to illustrate the utility of the model, it could be developed and applied to other continental shelf areas for which the necessary physical and biological data exist. The distribution of habitat sensitivities was compared with the distribution of international bottom-trawling activity. We used the predictions of sensitivity to compare the relative environmental costs of trawling in habitats with different sensitivities and show how modifications to the existing distribution and intensity of trawling disturbance would affect the aggregate impacts of trawling. Our methods provide clear quantitative guidance for assessing the outcome, in terms of environmental costs and benefits (changes in the total biomass and production of benthic invertebrate communities), of different management options designed to support implementation of an EAF.
- Top of page
- Supporting Information
The values of sensitivity for biomass (SB) and production (SP) increased with recovery time for production (TP) and biomass (TB) (Fig. 1). The spatial distributions of SB and SP in the North Sea (Fig. 2) showed that they tend to be highest in the north-west North Sea and Oyster Ground area, and many areas with relatively high or low SB and SP are broadly the same in their spatial distribution. One deviation from this general pattern occurred in areas where SB > 0·8 and there was slow recovery but low absolute biomass. This occurred in a narrow band from north-east England (Flamborough head) to the central coast of the Netherlands (IJmuiden).
The largest proportions (> 40%) of the study area had sensitivities in the ranges SP = 0·2–0·4 and SB = 0·6–0·8 (Table 2). The environmental characteristics associated with different values of S are given in Appendix S1 (see the supplementary material). Trawling disturbance was generally higher in areas of lower sensitivity. Frequently trawled areas (> 1 year−1) were uncommon and tended to occur primarily in areas of lower sensitivity (Fig. 3). However, some highly trawled areas corresponded with higher sensitivity (> 0·6). Thus trawling impacts would be expected to be much reduced if effort was reallocated to less sensitive habitats.
Table 2. Fraction of the area of North Sea seabed habitats assigned to sensitivity ranges (S). The total area of habitat was 125 000 km2
|S||Fraction of area (%) |
Figure 3. Trawling frequencies in habitats with different sensitivities S (0, low sensitivity to 1, high sensitivity).
Download figure to PowerPoint
The magnitude of predicted trawling impacts on production and biomass were correlated with the SP and SB index (Fig. 4). These relationships were used to compare the relative effects of trawling on P and B in habitats with different sensitivities. For example, trawling 2 year−1 in a habitat with SP = 0·2 is equivalent to trawling once every three years in a habitat with SP = 0·4. Trawling 5 year−1 in habitat with SP = 0·2 is equivalent to trawling four times every 10 years in habitat with SP = 0·8. A trawling frequency of 5 year−1 in the least-sensitive habitat had the same ecological effect as trawling 0·3 year−1 in the most-sensitive habitat, based on production. At a trawling intensity of 1 year−1, trawling 1 km2 of habitat with SP = 0·5 has an impact on production equivalent to trawling 3 km2 of habitat with SP = 0·2. The ecological impact increased with SP, while the relationship between SB and ecological impact was dome-shaped.
Figure 4. Isoclines of equivalent reduction in benthic production (a) and biomass (b) as a result of trawling disturbance in relation to habitat sensitivity. The lines connect points of equal impact (mg m−2) on production or biomass. Data have been smoothed using a Loess smoother.
Download figure to PowerPoint
The ecological impact of trawling in the North Sea could be reduced substantially by allocating bottom-trawling to the least-sensitive habitats, even without reducing the overall trawling effort (Table 3). If trawling effort was increased, ecological damage could be kept at or below current levels (scenarios 2–3) by distributing trawling effort to the least-sensitive habitats. If trawling effort was kept at current levels, ecological damage could be reduced to 36% (production) or 25% (biomass) of current levels by redistributing trawling to the least-sensitive habitats (scenario 4). When trawling effort was concentrated in the most-sensitive habitats, the ecological impact changed to 135% (production) or 105% (biomass) of current levels (scenario 5).
Table 3. The effect of different management scenarios on trawling effort and the ecological impact of trawling in the North Sea. Values shown in italics were set as constants, other values were derived from these. As cells are trawled at a frequency of 1 year−1, the trawling frequency is equal to the fraction of the area trawled. P, production; B, biomass
|Scenario||Trawling frequency (mean, year−1)||Ecological impact (relative to current levels) (%) |
|1 Ecological impact reduced to 25% bychanging trawling effort level and distribution||0·37||0·43|| 25|| 25|
|2 Ecological impact reduced to 50% bychanging trawling effort level and distribution||0·50||0·57|| 50|| 50|
|3 Ecological impact kept equal by changingtrawling effort level and distribution||0·68||0·79||100||100|
|4 Effort 100%, minimize ecological impact||0·43||0·43|| 36|| 25|
|5 Effort 100%, maximize ecological impact||0·43||0·43||135||105|
- Top of page
- Supporting Information
We have described a quantitative, field-data validated and repeatable method for assessing the sensitivity of seabed habitats to physical disturbance and delineating and mapping habitat sensitivity at large spatial scales. Although this approach was applied to North Sea seabed communities, and to assess sensitivity in relation to the additional physical disturbance associated with bottom-trawling, the model used to predict the recovery of production or biomass could readily be parameterized with different mortality functions to look at the effects of other types of disturbance, such as aggregate extraction, or modified to examine the effects of other types of fishing gear (Kaiser et al. 2006). The capacity to make quantitative predictions of sensitivity enables managers to take account of habitat sensitivity explicitly when establishing spatial management plans. In the context of fisheries management, this will support the identification and selection of fishing grounds and closed areas (both representative and highly sensitive), the comparison of management options that might reduce the overall environmental impacts of fishing, and future steps towards environmental impact assessment in advance of fishery development. Currently, there exists a large range of policy commitments to maintain and promote the sustainability of the exploitation of marine living resources, for example to ‘promote sustainable fisheries consistent with a diverse and resilient marine environment’ (DARDNI et al. 2005) and to ‘maintain the productivity of important and vulnerable marine and coastal areas’ (WSSD 2002).
The proposed method for classifying habitat sensitivity provides a logical and ecologically meaningful basis for assessing and improving the effectiveness of management. Thus, by directing trawling to the least-sensitive habitats, ecological damage can be reduced without reducing total trawling effort. Coupled with knowledge of the distribution of target fish species, our approach may help identify fishing grounds where an appropriate balance between catch rates and environmental impacts might be achieved (Jennings 2000).
Estimates of habitat sensitivity may be used to assign fishing access rights based on the ecological impact of trawling (Holland & Schnier 2006) and for identifying ‘ecologically expensive’ interactions between trawling and the environment, in which short-term financial gains result in long-term environmental degradation. In future studies, ecological impacts in terms of lost production or biomass might be expressed in terms of monetary value (Costanza et al. 1997; Chee 2004). Trawlers could then be charged differentially for access rights to fishing grounds (habitats), depending on their sensitivity (Holland & Schnier 2006). This is a form of ecosystem rent that assigns a monetary value to each habitat according to its sensitivity.
The choice of fishing location by trawlers would thus be expected to depend on the spatial distribution of ecosystem rent in relation to fishing profits. To predict these choices of fishing location, it would be necessary to model how the behaviour of fishermen is affected by the balance between the charges and expected fishing profits. Such models are relatively straightforward extensions of existing models that predict the spatial distribution of effort from the predicted behaviour of fishers (Hutton et al. 2004; Hiddink et al. 2006a).
Holland & Schnier (2006) recently proposed an individual habitat quota (IHQ) system that uses economic incentives to encourage habitat conservation. Individual quotas of habitat impact units (HIU) would be distributed to fishers, with a cumulative quota set for an area to maintain a target habitat ‘stock’. The use of HIU by fishers thus reflects the amount of habitat damage that they cause by trawling. Holland & Schnier (2006) assumed that there was no spatial variation in habitat sensitivity to trawling, and that the absolute amount of trawling damage caused by trawling decreased exponentially with the number of previous trawling events (trawling history). Their approach could be developed further using our measures of the sensitivity of benthic communities, as the charges for trawling (use of HIU) would depend both on trawling history and on the sensitivity of the targeted area. The obvious advantage of this approach is that it becomes economically less attractive to fish in the most-sensitive areas and fisheries targeting relatively mobile fish would probably focus on areas where costs were lower (Holland & Schnier 2006). This approach could be developed further into a management system where gear types that fished on a habitat of given sensitivity would be charged differentially according to their relative impact and frequency of use. The impacts of different gears on different habitats could be assessed based on a meta-analysis of studies of the impacts of different gear types on different habitats (Kaiser et al. 2006).
Some problems have to be acknowledged with regard to ecosystem impact quotas. If such quotas are non-transferable, all the existing problems that result from non-transferability of fishing quotas (Arnason 2005) are likely to affect habitat impact quotas. Several fleets using different gears operate in the North Sea, with varying impact on the environment. Fleets fishing high-priced species would have an inherent advantage if ecosystem impact quotas were used, all other things being equal, as their profits in an area where they have to pay for a moderate impact may still be larger than the profits of other fisheries in areas where they have a low impact. Thus, if HIU are too cheap or too expensive relative to the profits of a fleet, HIU are unlikely to help ecosystem conservation. The risk also exists that quotas are monopolized for purposes other than the managers may have intended (e.g. by environmental organizations). Therefore, implementing HIU will not be simple.
Our analyses demonstrate that the allocation of bottom-trawling effort to less-sensitive habitats could lead to substantial reductions in the ecological impact of trawling, even without reductions in overall trawling effort. In this analysis, we assumed that all areas were suitable for trawling. However, benthic invertebrate production is typically low in areas that are least sensitive to trawling impacts, and it is possible that fish abundance, catch rates and total catch would also be low in these areas (Hoines & Bergstad 1999). If catch rates or total catch and habitat sensitivity were directly and positively related, then the redistribution of trawling would not provide any environmental benefits over and above a simple reduction of the total trawling effort. However, relationships between sensitivity and catch rates are unlikely to be direct and positive for at least three reasons. First, most North Sea species migrate, and while the areas occupied in some seasons may be sensitive and/or productive there will be times when, and places where, the fish can be caught over habitat of lower sensitivity. Secondly, aggregations of fish that are targeted by fishers can be very concentrated (particularly given the low abundance of many North Sea stocks) and the concentrations may reflect other drivers of habitat suitability, such as temperature, as well as the physical structure and productivity of the benthic habitat (Shepherd & Litvak 2004). Thirdly, trawlers often tend to trawl along tows that are recorded in their navigation systems and this results in patterns of effort that persist over many years (Auster & Langton 1999; Holland & Sutinen 2000). Additionally, at trawling intensities > 1 year−1, the absolute differences between production and biomass at different levels of SP and SB are relatively small. Thus it seems unlikely that catch rates and habitat sensitivity will usually be directly and positively related, and thus it would be possible to redirect fleets to areas where the reductions in catch rates would be small in relation to the ecological benefits realized. Future studies should look in detail at the trade-offs between catch rate and ecological impact, and how they might vary seasonally, based on catch rate data and knowledge of fish migration.
The principal limitations of our method are that habitat sensitivity is measured in terms of production and biomass recovery rather than in terms of other attributes, such as diversity, that may be regarded as more important in a policy context. Moreover, the connectivity among habitats, habitat diversity and patch size were not accounted for in our model. Based on comprehensive validation, and the open nature of the North Sea marine environment, we consider these to be acceptable simplifications (Hiddink et al. 2006b). For other areas, the effects of disregarding connectivity and other processes would have to be assessed. The implementation of the management scenarios was also simplified by assuming that all areas are either not trawled or trawled once a year. Given that the reduction in production and biomass at low trawling intensities is large, while further increasing the trawling intensity in areas where trawling intensity is already high only has a small impact, a concentrated rather than a homogeneous distribution of a given level of effort will have a lower ecological impact. For all sensitivities, the relationship between trawling intensity and ecological impact flattens off at trawling intensities > 1 year−1. This explains the seemingly counterintuitive result of scenario 5, where concentrating fishing effort in the most-sensitive habitats still only led to a very small increase in the ecological impact of trawling on biomass (104%). To avoid managers forming an unduly optimistic impression of the result of their actions, an assessment of the impacts when effort was not constrained to 1 year−1 would be necessary.
Concentrating trawling effort, regardless of the sensitivity of the habitat in which effort is concentrated, will reduce the aggregate ecological impact of trawling on that habitat. Trawlers could be encouraged to tow in defined narrow lanes (e.g. < 1 km wide) if it was essential to trawl in some areas of a sensitive habitat to maintain catch rates of some species. This would leave most of the habitat unimpacted, while the narrow lanes would facilitate immigration of commercial fish into the trawl path. The feasibility of such an approach depends on fish production rates in the untrawled areas, the movement of fish in relation to the trawl lanes, and the capacity of managers to monitor and regulate trawling activities reliably at this scale. With the advent of VMS, the spatial management of trawling at scales of a few kilometres is increasingly realistic.
Notwithstanding the limitations of our approach, we have shown that models of recovery time can be used to estimate habitat sensitivity and that the results can be used to map sensitive habitats, compare the sensitivity of habitats on a common scale and select spatial management strategies that minimize human impacts. Our methods do not rely on expert judgement and/or scoring systems that have commonly been adopted elsewhere (Zacharias & Gregr 2005) and, at least in the study area, are validated, repeatable and applicable at the scale of management.